Mon. Not. R. Astron. Soc. 396, 849859 (2009) doi:10.1111/j.1365-2966.2009.14526.

x
The impact of dust on the scaling properties of galaxy clusters
Antonio C. da Silva,
1,2
Andrea Catalano,
3
Ludovic Montier,
4
Etienne Pointecouteau,
4
Joseph Lanoux
4
and Martin Giard
4
1
Centro de Astrosica da Universidade do Porto, Rua das Estrelas, 4150-762 Porto, Portugal
2
Institut dAstrophysique Spatiale, Bat 121, Universit e Paris-Sud, 91405 Orsay, France
3
Observatoire de Paris LERMA, 61 avenue de lObservatoire, 75014 Paris, France
4
Universit e de Toulouse, CNRS, UMR 5187 CESR, 9 av. du Colonel Roche, 31400 Toulouse, France
Accepted 2009 January 19. Received 2008 December 23; in original form 2008 August 14
ABSTRACT
We investigate the effect of dust on the scaling properties of galaxy clusters based on hy-
drodynamic N-body simulations of structure formation. We have simulated ve dust models
plus radiative cooling and adiabatic models using the same initial conditions for all runs. The
numerical implementation of dust was based on the analytical computations of Montier &
Giard. We set up dust simulations to cover different combinations of dust parameters that
make evident the effects of size and abundance of dust grains. Comparing our radiative plus
dust cooling runs with a purely radiative cooling simulation, we nd that dust has an impact
on cluster scaling relations. It mainly affects the normalization of the scalings (and their evo-
lution), whereas it introduces no signicant differences in their slopes. The strength of the
effect critically depends on the dust abundance and grain size parameters as well as on the
cluster scaling. Indeed, cooling due to dust is effective in the cluster regime and has a stronger
effect on the baryon driven statistical properties of clusters such as L
X
M, YM, SM scaling
relations. Major differences, relative to the radiative cooling model, are as high as 25 per cent
for the L
X
M normalization, and about 10 per cent for the YM and SM normalizations at
redshift zero. On the other hand, we nd that dust has almost no impact on the dark matter
driven T
mw
M scaling relation. The effects are found to be dependent in equal parts on both
dust abundances and grain size distributions for the scalings investigated in this paper. Higher
dust abundances and smaller grain sizes cause larger departures from the radiative cooling
(i.e. with no dust) model.
Key words: methods: numerical galaxies: clusters: general large-scale structure of
Universe.
1 I NTRODUCTI ON
From the rst stages of star and galaxy formation, non-gravitational
processes drive together with gravitation the formation and the evo-
lution of structures. The complex physics they involve rule the
baryonic component within clusters of galaxies, and in a more gen-
eral context within the intergalactic medium (IGM; see the review
by Voit (2005) and references therein). The study of these processes
is the key to our understanding of the formation and evolution of
the large-scale structure of the Universe. Indeed, understanding how
their heating and cooling abilities affect the thermodynamics of the
IGM at large scales and high redshifts, and thus that of the intra-
cluster medium (ICM) once the gas gets accreted on to massive
haloes, is a major question still to be answered. The continuous

E-mail: asilva@astro.up.pt
accretion and the merger events through which a halo is assembled
lead to a constant interaction of the IGMgas with the evolving galac-
tic component. Within denser environments, like clusters, feedback
provided by active galactic nuclei (AGN) balances the gas cooling
(see, e.g., Cattaneo & Teyssier 2007; McNamara & Nulsen 2007;
Conroy &Ostriker 2008 for a review). Also, fromhigh redshifts, the
rate of supernovae drives the strength of the galactic winds and thus
the amount of material that ends up being ejected within the IGM
and the ICM (see Loewenstein 2006). These ejecta are then mixed
in the environment by the action of the surrounding gravitational
potential and the dynamics of cluster galaxies within.
For a long time, X-ray observations have shown the abundant
presence of heavy elements within the ICM (see e.g., Sarazin 1988;
Arnaud 2005). Physical processes like ram-pressure stripping, AGN
interaction with the ICM, galaxygalaxy interaction or mergers are
scrutinized within analytical models and numerical simulations in
order to explain the presence of metals (see, e.g., Domainko et al.
C _
2009 The Authors. Journal compilation
C _
2009 RAS
850 A. C. da Silva et al.
2006; Kapferer et al. 2006; Moll et al. 2007). Moreover, it is obvious
that the process of tearing of material from galaxies leads not only
to the enrichment of the ICM/IGM in metal, but in gas, stars and
dust as well.
Recent work on numerical simulations (Murante et al. 2004,
2007; Conroy, Wechsler & Kravtsov 2007) has stressed the role of
hierarchical building of structures in enriching the ICM with stars
in a consistent way with the observed amount of ICM globular
clusters, and ICM light. Indeed, the overall light coming from stars
in between cluster galaxies represents an important fraction of the
total cluster light: for instance, Krick & Bernstein (2007) measured
622 per cent from a sample of 10 clusters. The effect of a diffuse
dust component within the IGM/ICM is less well known. A few
observational studies with the ISO and Spitzer satellites have tried
without success to detect the signature of such a component (Stickel
et al. 1998, 2002; Bai et al. 2006, 2007). More successfully, Montier
& Giard (2005) have obtained a statistical detection, via a stacking
analysis, of the overall infrared (IR) emission coming from clusters
of galaxies. However, they were not able to disentangle the IRsignal
from dusty cluster galaxies from a possible ICM dust component.
On the other hand, froma theoretical point of viewa fewworks have
looked at the effect of dust on the ICM (Popescu et al. 2000) or in
conjunction with the enrichment of the ICMin metals (Aguirre et al.
2001). However, the effect of dust on a ICM/IGM-type thermalized
plasma has been formalized by Montier & Giard (2004). These
authors have computed the cooling function of dust taking into
account the energetic budget for dust. They have shown the ability
of dust to be a non-negligible cooling/heating vector depending on
the physical properties of the environment.
Dust thus comes, within the ICM/IGM, as an added source of
non-gravitational physics that can potentially inuence the forma-
tion and the evolution of large-scale structure in a signicant way.
Indeed, since a redshift of z . 25 during which the star formation
activity reached its maximum in the cosmic history, large amounts
of dust have been produced and thus ejected out of galaxies due to
violent galactic winds into the IGM (Springel & Hernquist 2003).
As this material is then accreted by the forming haloes, one can
wonder about the impact produced by dust on the overall properties
of clusters of galaxies once assembled and thermalized. In a hierar-
chical Universe, the population of clusters is self-similar, thus it is
expected to present well-dened structural and scaling properties.
However, to date, it is common knowledge that the observed prop-
erties deviate from the prediction by a purely gravitational model
(see, Arnaud 2005; Voit 2005 for reviews). It is thus important to
address the issue of the impact of dust on the statistical properties
of structures such as clusters of galaxies, in the same way as is done
for AGNs, supernovae, stripping or mergers.
In order to tackle this question, we have put into place the rst
N-body numerical simulations of hierarchical structure formation
implementing the cooling effect of dust according to the dust nature
and abundance. In this paper, we present the rst results of this
work focusing on the scale of galaxy clusters, and more specically
on their scaling properties. The paper is organized as follows: we
start by presenting the physical dust model and how it is imple-
mented in the numerical simulation code. In Section 3, we describe
the numerical simulations and the various runs (i.e. model) that
have been tested. From these simulations, our analysis concerns
the galaxy cluster scale, and focuses on the impact of the pres-
ence of dust on the scaling relation of clusters. In Section 4, we
present our results on the TM, SM, YM and L
X
M relations.
The derived results are presented in Section 5 and discussed in
Section 6.
2 THE DUST MODEL
In our numerical simulations, the implementation of the physical
effect of the dust grains is based on the computation by Montier &
Giard (2004) of the dust heating/cooling function. In this work, we
decided to limit our implementation to the dust cooling effect only.
Indeed, the goal of this paper is to study the effect of dust at galaxy
cluster scales. The heating by the dust grains is mainly effective at
low temperatures (i.e. T
e
<10
5
K) and is a localized effect strongly
dependent on the ultraviolet (UV) radiation eld. Our numerical
simulations (see Sections 3 and 6.2) do not directly implement this
level of physics.
The dust grains in a thermal plasma with 10
6
<T <10
9
Kare de-
stroyed by thermal sputtering, the efciency of which was quantied
by Draine & Salpeter (1979, see their equation 44). The sputtering
time depends on the column density and on the grain size. For grain
sizes ranging from 0.001 to 0.5 m, and an optically thin plasma
(n 10
3
atomcm
3
), the dust lifetime ranges from10
6
yr for small
grains up to 10
9
yr for big grains. These lifetimes are therefore large
enough for the cooling by dust in the IGM/ICM to be considered.
Evidently, it is also strongly linked to the injection rate of dust, thus
to the physical mechanism that can bring and spread dust in the
IGM/ICM. In the present work, as a rst step, we limit ourselves
to a basic approach of the implementation of the dust cooling in
numerical simulations of structure formation (see Section 2.3). We
have not implemented the physical processes of dust creation (i.e.
injection in the ICM) and destruction (i.e. sputtering), which de
facto put some intrinsic limitations on the derived results (see the
discussion in Section 6.2).
Our implementation of the dust cooling power is based on the
model of Montier &Giard (2004). We recall belowthe main aspects
of this model and describe the practical implementation within the
N-body simulations.
2.1 The dust cooling function
Dust grains within a thermal gas such as the ICM or the IGM can be
either a heating or a cooling vector depending on the physical state of
the surrounding gas and on the radiative environment. Heating can
occur via the photoelectric effect if the stellar radiation eld (stars
and/or QSOs) is strong enough (Weingartner, Draine & Barr 2006
and references therein). Indeed, the binding energies of electrons in
the dust grains are small, thus allowing electrons to be more easily
photo-detached than in the case of a free atom or a molecule. On
the other hand, the cooling by dust occurs through re-radiation in
the IR of the collisional energy deposited on grains by impinging
free electrons of the ICM/IGM.
1
Montier & Giard (2004) have computed the balance of heating
and cooling by dust with respect to the dust abundance: cooling by
dust dominates at high temperatures in the hot IGM of virialized
structures (i.e. clusters of galaxies), and heating by dust dominates
in low-temperature plasma under high radiation uxes such as in
the proximity of quasars. The details, of course, depend on the local
physical parameters such as the grain size and the gas density.
Assuming local thermal equilibrium for the dust, the overall bal-
ance between heating and cooling in the dust grains can be written
as follows:

g
(a, T
d
) = H
g
coll
(a, T
e
, n
e
), (1)
1
In the galactic medium, the cooling occurs through re-radiation of the
power absorbed in the UV and visible range.
C _
2009 The Authors. Journal compilation
C _
2009 RAS, MNRAS 396, 849859
The impact of dust on cluster scaling laws 851
with H
coll
being the collisional heating function of the grain and
the cooling function due to thermal radiation of dust. a is the
grain size, T
e
and n
e
are, respectively, the electronic temperature
and density of the medium and T
d
is the dust grain temperature.
The heating of the dust grain was taken from Dwek (1981) and
can be expressed in a general way as
H
g
coll
(a, T
e
, n
e
) n
e
a

e
, (2)
where the values of and are dependent on the value of the ratio
a
2/3
/T
e
.
The relevant dust parameters affecting the cooling function are
the grain size and the metallicity. Indeed, the smaller the grains and
the higher the metallicity, the higher is the cooling power of the
dust. Thus, the total cooling function due to a population of dust
grains can be expressed as a function of these two parameters as
(a, T
d
) =


g
(a, T
d
)
dN(a, Z, V)
dV da dZ
dV da dZ, (3)
where dN(a, Z, V)/dV da dZ is the differential number of dust
grains per size, metallicity and volume element.
Cooling by dust happens to increase with the square root of the
gas density, whereas heating by dust is proportional to the density.
As stressed by Montier & Giard (2004), cooling by dust is more
efcient within the temperature range of 10
6
< T < 10
8
K (i.e.
0.1 < kT < 10 keV), which is typically the IGM and ICM thermal
condition.
We redirect readers to Montier & Giard (2004) for a full descrip-
tion of the dust model, and a comprehensive physical analysis of
the effect of dust in a optically thin plasma.
2.2 The dust abundance
The abundance of dust is a key ingredient to properly weight in our
implementation. Observations indicate that dust represents only a
tiny fraction of the baryonic matter: M
dust
/M
gas
0.01 in our Milky
Way (Dwek, Rephaeli &Mather 1990), and this is possibly lower by
a factor of 1001000 in the ICM: M
dust
/M
gas
=10
5
10
4
(Popescu
et al. 2000; Aguirre et al. 2001). We dened the abundance of dust
as the ratio of the dust mass with respect to the gas mass:
Z
d
=
M
dust
M
gas
= f
d
Z
Z

Z
d
, (4)
where Z is the metallicity in units of solar metallicity, Z
d
=0.0075
is the solar dust abundance, that is the dust-to-gas mass ratio in the
solar vicinity (Dwek et al. 1990), and f
d
is the abundance of dust in
the ICM in units of solar dust abundance.
Dust enrichment occurs via the feedback of galaxy formation
and evolution in the ICM through interaction, stripping, mergers,
galactic winds and AGN outbursts. At all redshifts, it is linked to
the star formation rate (SFR) which drives the production of dust
in cluster galaxies. However, in our hydrodynamic simulations (see
Section 3), the SFR is not physically modelled, but it is inferred
by the cooling state of the gas particles within the simulations:
gas particles below a given threshold of temperature and above a
given threshold of density are considered as collisionless matter,
forming stars and galaxies (see Section 3). In order to tackle this
problem, we choose to directly link the dust abundance to the metal
abundance using equation (4). Therefore, the dust distribution in
our simulations mimics the metal distribution.
2.3 Implementation in the N-body simulations
Fromthe equations presented in the previous sections, we computed
the dust cooling function according to the embedding medium tem-
Figure 1. Cooling functions implemented in the numerical simulations.
The top panel shows the dependence of the dust model D1 (f
d
= 0.1 and
a = 10
3
m) on metallicity (and temperature) whereas the bottom panel
shows different dust models at the same metallicity Z/Z

=0.33 (see text).

Black, blue and red curves are the total cooling functions, radiative cooling
of the gas from Sutherland & Dopita (1993) and dust cooling functions,
respectively.
perature and (global) metallicity. In simulations, once the metallicity
and temperature are known, a and f
d
are the only two parameters
driving the dust cooling rate [i.e. (a, Z) = (a, f
d
)]. In the top
panel of Fig. 1, we present dust cooling rates (red lines) for f
d
=
0.1 and a = 10
3
m (model D1, see below) at different values of
metallicity. The blue and black lines are the radiative cooling rates
from Sutherland & Dopita (1993) and the total (i.e. radiative plus
dust cooling) rate, respectively.
Together with an adiabatic run (i.e. model A) and a standard
radiative run (model C see Section 3 for further details), we ran
a total of ve runs implementing various populations of grains (i.e.
named D1 to D5) characterized by their size and dust-to-metal mass
ratio.
(i) We tested three types of sizes: two xed grain sizes with a =
10
3
and =0.5 m, respectively, labelled small and big. The third
assumes for the IGM dust grains a distribution in sizes as dened
by Mathis, Rumpl & Nordsieck (1977) for the galactic dust: N(a)
a
3.5
within the size interval of [0.001, 0.5] m. It is hereafter
referred to as the MRN distribution.
C _
2009 The Authors. Journal compilation
C _
2009 RAS, MNRAS 396, 849859
852 A. C. da Silva et al.
Table 1. Simulation parameters: f
d
, dust-to-metal mass ratios (see equa-
tion 4), grain sizes and number of time-steps taken by simulation runs to
evolve from z = 49 to 0. Cosmological and simulation parameters were set
the same in all simulations, as follows: = 0.3,

= 0.7,
b
= 0.0486,

8
= 0.9, h = 0.7, boxsize L = 100 h
1
Mpc and number of baryonic and
dark matter particles, N = 4 096 000.
Run Physics f
d
Grain size N
steps
A Adiabatic (no dust) 2569
C Cooling (no dust) 2633
D1 Cooling with dust 0.100 Small 2944
D2 Cooling with dust 0.100 MRN 2920
D3 Cooling with dust 0.100 Big 2886
D4 Cooling with dust 0.010 MRN 2698
D5 Cooling with dust 0.001 MRN 2633
(ii) We investigate three values of f
d
: 0.001, 0.01 and 0.1. The two
extreme values roughly bracket the current theoretical and obser-
vational constraints on dust abundance in the ICM/IGM (i.e. 10
5
and 10
3
in terms of dust-to-gas mass ratio). Different observations
published in the literature give different constraints for this parame-
ter. Z
d
=5 10
4
(i.e. f
d
=0.2) is the upper limit derived by Giard
et al. (2008), if all the IR luminosity they have statistically detected
towards galaxy clusters is produced by thermal emission from intr-
acluster dust. This is very close to the inferred upper limit of Z
d
=
2 10
4
(i.e. f
d
= 0.08) that we can derive from the non-detection
of the statistical reddening of background galaxies behind 458 RCS
clusters of Muller et al. (2008) (the dust mass within the central Mpc
is found to be less than 8 10
9
M

). A very similar work by Bovy,

Hogg & Moustakas (2008) on Sloan Digital Sky Survey (SDSS)
clusters at z .0.05 led to a colour excess upper limit of E(B V) <
3 10
3
mag on scales of 12 Mpc (i.e. an encompassed dust mass
of 10
8
M

). This value matches the results of Chelouche, Koester

& Bowen (2007), who averaged the reddening of QSOs behind the
SDSS clusters as a function of the impact parameter on the clusters.
These authors measured an average colour excess of E(g i)) .3
10
3
mag. This detection is statistically signicant for large im-
pact parameters, i.e. 1 R 6 7 R
200
(where R
200
.1 Mpc). If
we extrapolate their measurements to the central Mpc of a cluster,
this extinction translates into a dust mass of M
dust
= 3 10
8
M

(see their equation 4). Compared with the corresponding gas mass
in the same volume, this leads to Z
d
. 10
5
(i.e. f
d
. 0.004). Our
chosen values of f
d
thus bracket the range of current observational
constraints.
Table 1 lists code names and simulation details of all runs used in
this work. In the case of models D1 to D5, simulation cooling rates
are given by the added effect of cooling due to dust and radiative gas
cooling. Total cooling functions are displayed (non-coloured lines)
in the bottom panel of Fig. 1 for each of these models at Z/Z

=
0.33. As the gure indicates, the effect of dust cooling is stronger
for models with higher dust-to-metal mass abundance parameters,
f
d
, and for smaller grain sizes (model D1). For low values of f
d
,
the impact of dust cooling is signicantly reduced. For example, in
the case of model D5, the contribution of dust to the total cooling
rate is negligible at Z/Z

=0.33 for all temperatures. Therefore, we

do not expect to obtain signicant differences between simulations
with these two models.
3 NUMERI CAL SI MULATI ONS
3.1 Simulation description
Simulations were carried out with the public code package HYDRA,
(Couchman, Thomas & Pearce 1995; Pearce & Couchman 1997),
an adaptive particleparticle/particlemesh (AP
3
M) (Couchman
1991) gravity solver with a formulation of smoothed particle hydro-
dynamics (SPH), see Thacker & Couchman (2000), that conserves
both entropy and energy. In simulations with cooling gas parti-
cles are allowed to cool using the method described in Thomas
& Couchman (1992) and the cooling rates presented in the previ-
ous section. At a given time-step, gas particles with overdensities
(relative to the critical density) larger than 10
4
and temperatures
below 1.2 10
4
K are converted into collisionless baryonic mat-
ter and no longer participate in the gas dynamical processes. The
gas metallicity is assumed to be a global quantity that evolves
with time as Z = 0.3(t/t
0
) Z

, where Z

is the solar metal-

licity and t/t
0
is the age of the Universe in units of the current
time.
All simulations were generated from the same initial conditions
snapshot, at z = 49. The initial density eld was constructed, using
N = 4 096 000 particles of baryonic and dark matter, perturbed
from a regular grid of xed comoving size L = 100 h
1
Mpc. We
assumed a cold dark matter cosmology with parameters =0.3,

= 0.7,
b
= 0.0486,
8
= 0.9, h = 0.7. The amplitude of the
matter power spectrum was normalized using
8
=0.9. The matter
power spectrum transfer function was computed using the BBKS
formula (Bardeen et al. 1986), with a shape parameter given by
the formula in Sugiyama (1995). With this choice of parameters,
the dark matter and baryon particle masses are 2.1 10
10
and 2.6
10
9
h
1
M

, respectively. The gravitational softening in physical

coordinates was 25 h
1
kpc below z = 1 and above this redshift
scaled as 50(1 z)
1
h
1
kpc.
We generated a total of seven simulation runs, listed in Table 1.
The rst two runs, which will be referred to hereafter as adiabatic
(or model A) and cooling (or model C) simulations, do not
include dust. Simulations 3 to 7 differ only in the dust model pa-
rameters assumed in each case, and will be referred to as dust
runs, and are labelled as D1 to D5 models (see Section 2.3 for
details on the dust models denition). This will allow us to investi-
gate the effects of the dust model parameters on our results. The last
column in the table gives the total number of time-steps required by
each simulation to arrive to redshift zero. For each run, we stored a
total of 78 snapshots in the redshift range 0 < z < 23.4. Individual
snapshots were dumped at redshift intervals that correspond to the
light travel time through the simulation box, i.e. simulation outputs
stack in redshift.
3.2 Catalogue construction
Cluster catalogues are generated from simulations using a modied
version of the Sussex extraction software developed by Thomas and
collaborators (Thomas et al. 1998; Pearce et al. 2000; Muanwong
et al. 2001). Briey, the cluster identication process starts with
the creation of a minimal-spanning tree of dark matter particles
which is then split into clumps using a maximum linking length
equal to 0.5
1/3
b
times the mean interparticle separation. Here,

b
is the contrast predicted by the spherical collapse model of a
virialized sphere (Eke, Navarro & Frenk 1998). A sphere is then
grown around the densest dark matter particle in each clump until
C _
2009 The Authors. Journal compilation
C _
2009 RAS, MNRAS 396, 849859
The impact of dust on cluster scaling laws 853
the enclosed mass veries
M

(< R

) =
4
3
R
3

crit
(z), (5)
where is a xed overdensity contrast,
crit
(z) = (3H
2
0
/8G)E
2
(z)
is the critical density and E(z) = H(z)/H
0
=

((1 z)
3

.
Cluster properties are then computed in a sphere of radius R
200
, i.e.
with = 200, for all objects found with more than 500 particles
of gas and dark matter. This means that our original catalogues
are complete in mass down to 1.18 10
13
h
1
M

. For the study

presented in this paper, we have trimmed our original catalogues to
exclude galaxy groups with masses belowM
lim
=5 10
13
h
1
M

.
In this way, the less massive object considered in the analysis is
resolved with a minimum of 2100 particles of both gas and dark
matter. Our catalogues at z =0 have at least 60 clusters with masses
above M
lim
. This number drops to about 20 clusters at z = 1.
Cluster properties investigated in this paper are the mass, M,
mass-weighted temperature, T
mw
and entropy, S (dened as S =
k
B
T/n
2/3
), integrated Compton parameter, Y (i.e. roughly the SZ
signal times the square of the angular diameter distance to the
cluster), and core excised (50 h
1
kpc) X-ray bolometric luminosity,
L
X
. These were computed in the catalogues according to their usual
denitions, see da Silva et al. (2004),
M =

k
m
k
, (6)
T
mw
=

i
m
i
T
i

i
m
i
, (7)
S =

i
m
i
k
B
T
i
n
2/3
i

i
m
i
, (8)
Y =
k
B

T
m
e
c
2
(1 X)
2m
H

i
m
i
T
i
, (9)
L
X
=

i
m
i

i

bol
(T
i
, Z)
(m
H
)
2
, (10)
where summations with index i are over hot (T
i
> 10
5
K) gas par-
ticles and the summation with index k is over all (baryon and dark
matter) particles within R
200
. Hot gas is assumed fully ionized.
The quantities m
i
, T
i
, n
i
and
i
are the mass, temperature, number
density and mass density of gas particles, respectively.
bol
is the
bolometric cooling function in Sutherland & Dopita (1993) and Z
is the gas metallicity. Other quantities are the Boltzmann constant,
k
B
, the Thomson cross-section,
T
, the electron mass at rest, m
e
,
the speed of light c, the hydrogen mass fraction, X = 0.76, the gas
mean molecular weight, , and the hydrogen atom mass, m
H
.
4 SCALI NG RELATI ONS
In this paper, we investigate the scalings of mass-weighted temper-
ature, T
mw
, entropy, S, integrated Compton parameter, Y, and core
excised X-ray bolometric luminosity, L
X
, with mass, M. Taking into
account equation (5) these cluster scaling relations can be expressed
as
T
mw
= A
TM
(M/M
0
)

TM
(1 z)

TM
E(z)
2/3
, (11)
S = A
SM
(M/M
0
)

YT
(1 z)

YT
E(z)
2/3
, (12)
Y = A
YT
(M/M
0
)

YM
(1 z)

YM
E(z)
2/5
, (13)
L
X
= A
LM
(M/M
0
)

LM
(1 z)

LM
E(z)
7/3
, (14)
where M
0
= 10
14
h
1
M

and the powers of the E(z) give the pre-

dicted evolution, extrapolated from the self-similar model (Kaiser
1986) of the scalings in each case. The quantities A, and are
the scaling normalization at z = 0, the power on the independent
variable, and the departures fromthe expected self-similar evolution
with redshift.
These scalings can be expressed in a condensed form,
y f (z) = y
0
(z) (x/x
0
)

, (15)
where y and x are cluster properties (e.g. T
mw
, M)
y
0
(z) = A(1 z)

, (16)
and f(z) is some xed power of the cosmological factor E(z). To
determine A, and for each scaling we use the method described
in da Silva et al. (2004) and Aghanim, da Silva & Nunes (2009).
To summarize, the method involves tting the simulated cluster
populations at each redshift with equations (15) and (16) written
in logarithmic form. First, we t the cluster distributions with a
straight line in logarithmic scale at all redshifts. If the logarithmic
slope remains approximately constant (i.e. shows no systematic
variations) within the redshift range of interest, we then set as the
best-tting value at z =0. Next, we repeat the tting procedure with
xed to (z = 0) to determine the scaling normalization factors
y
0
(z). This avoids unwanted correlations between and y
0
(z). The
rms dispersion of the t is also computed at each redshift according
to the formula

log y
/ =

1
N

i
[log(y
/
i
/y
/
)]
2
, (17)
where y
/
= yf (see equation 15) and y
/
i
are individual data points.
Finally, we perform a linear t of the normalization factors with
redshift in logarithmic scale, see equation (16), to determine the
parameters A and .
We note that above z = 1.5 the number of clusters in our cata-
logues decreases typically below 10, hence we do not t the scaling
relations above this redshift value.
5 RESULTS
5.1 Scaling relations at z = 0
In this section we present cluster scaling relations obtained from
simulations at redshift zero. We investigate the four scalings pre-
sented in Section 4 for all models under investigation.
Fig. 2 shows the T
mw
M (top left panel) S
mw
M, (top right panel),
YM (bottom left panel) and L
X
M (bottom right panel) scalings,
with all quantities computed within R
200
. In each case, the main plot
shows the cluster distributions for the C (triangles), D4 (diamonds),
D2 (lled circles) and D1 (crosses) simulations, whereas the em-
bedded plot presents the power-law best-tting lines (solid, triple
dotdashed, dashed and dotdashed for C, D4, D2 and D1 models,
respectively) obtained in each case, colour coded in the same way
as the cluster distributions. Here, we have chosen to display the
dust models that allow us to assess the effect of dust parameters
individually. For example, the dust models in runs D4 and D2 only
differ by the dust-to-metal mass ratio parameter, whereas models
D2 and D1 have different grain sizes but the same f
d
. The shaded
grey areas in the embedded plots give the rms dispersion of the t for
the cooling (C) model. The dispersions obtained for the other mod-
els have similar amplitudes to the C case. The scalings of entropy
C _
2009 The Authors. Journal compilation
C _
2009 RAS, MNRAS 396, 849859
854 A. C. da Silva et al.
Figure 2. Cluster scalings at redshift zero for T
mw
M (top left panel), SM, (top right panel), YM (bottom left panel) and L
X
M (bottom right panel).
Displayed quantities are computed within R
200
, the radius where the mean cluster density is 200 times larger than the critical density. Blue colour and triangles
stand for the cooling (C) run, cyan and diamonds are for the D4 run, yellow and lled circles are for clusters in the D2 run, and red and crosses are for the D1
run. The lines in the embedded plots are the best-tting lines to the cluster distributions and the shaded areas are the t rms dispersions for the C model, for
each scaling.
and X-ray luminosity with mass show larger dispersions because
they are more sensitive to the gas physical properties (density and
temperature) in the inner parts of clusters than the mass-weighted
temperature and Y versus mass relations which are tightly correlated
with mass.
An inspection of Fig. 2 allows us to conclude that the cluster scal-
ings laws studied here are sensitive to the underlying dust model, and
in particular to models where the dust cooling is stronger (model
D1 and D2). The differences are more evident in the SM and
L
X
M scalings, but are also visible, to a lower extent, in the
T
mw
M and YM relations. Generally, the inclusion of dust tends
to increase temperature and entropy because the additional cooling
increases the formation of collisionless (star-forming) material, see
Fig. 3, leaving the remaining particles in the gas phase with higher
mean temperatures and entropies. The decrease of Y and X-ray lu-
minosities reects the effect of lowering the hot-gas fraction and
density due to dust cooling. These effects dominate over the effect
of increasing the temperature.
In fact a closer inspection of Fig. 2 indicates that differences
for the same cluster in different models (note that all simulations
have the same initial conditions so a cluster-to-cluster comparison
can be made) reect the differences of intensity between cooling
functions presented in Fig. 1. For example, the differences between
models D4 and C are clearly small as one would expect from the
small differences between cooling functions displayed in the bottom
panel of Fig. 1. Another interesting example is that an increase of
one order of magnitude in f
d
from D4 to D2 seems to cause a
stronger impact in the properties of the most massive clusters than
the differences arising from changing the dust grain sizes from
D2 to D1. Again this reects the differences between the cooling
functions, which in the latter case are smaller at higher temperatures
(see bottom panel of Fig. 1).
Away of quantifying the effect of dust is to look at the best-tting
slope, , and normalization, log A, parameters of these scalings
which are presented in Table 2 for all cooling models considered
in this paper. We nd that tting parameters are quite similar for
C _
2009 The Authors. Journal compilation
C _
2009 RAS, MNRAS 396, 849859
The impact of dust on cluster scaling laws 855
Figure 3. Condensed baryon mass fraction (relative to the total mass) versus
mass within R
200
in clusters from simulations C, D4, D2 and D1 at redshift
zero. Symbols are the same as in Fig. 2.
models C, D5 and D4 whereas models with high dust abundances
provide the strongest variations of the tting parameters, particu-
larly for the normalizations. In several cases, differences are larger
than the (statistical) best-tting errors, particularly for the D1 and
D2 models. We also investigated scalings at redshift zero for the A
(adiabatic) model and found they were consistent with self-similar
predictions. As expected, the results obtained for the adiabatic and
cooling models are in very good agreement with the ndings of
da Silva et al. (2004) and Aghanim et al. (2009) which use similar
simulation parameters and cosmology.
We end this section with a discussion on the amount of condensed
baryons (star-forming material) that forms in our simulated clus-
ter populations. Fig. 3 shows the effect of dust on the fraction of
condensed baryons (mass of condensed baryons relative to the total
Table 2. Best-tting values of the parameters , log A and as well as their respective 1 errors. These values are valid within the redshift
range 0 < z < 1.
Model C Model D5 Model D4 Model D3 Model D2 Model D1
T
mw
M

TM
0.61 0.02 0.61 0.02 0.61 0.02 0.62 0.02 0.63 0.02 0.63 0.02
log A
TM
0.195 0.002 0.195 0.003 0.196 0.002 0.197 0.003 0.201 0.002 0.204 0.002

TM
0.14 0.01 0.14 0.01 0.14 0.01 0.15 0.01 0.16 0.01 0.16 0.01
SM

SM
0.55 0.03 0.54 0.03 0.54 0.03 0.56 0.03 0.55 0.02 0.54 0.02
log A
SM
2.443 0.002 2.444 0.002 2.445 0.002 2.451 0.002 2.468 0.002 2.488 0.02

SM
0.33 0.01 0.34 0.01 0.34 0.01 0.36 0.01 0.40 0.01 0.42 0.01
YM

YM
1.74 0.03 1.72 0.03 1.73 0.03 1.72 0.02 1.74 0.02 1.76 0.02
log A
YM
5.909 0.002 5.907 0.002 5.910 0.002 5.914 0.002 5.933 0.002 5.957 0.002

YM
0.12 0.01 0.11 0.01 0.13 0.02 0.13 0.01 0.17 0.01 0.21 0.01
L
X
M

LM
1.69 0.07 1.68 0.07 1.65 0.07 1.61 0.08 1.67 0.05 1.67 0.05
log A
LM
3.330 0.006 3.334 0.006 3.333 0.005 3.323 0.005 3.265 0.005 3.207 0.004

LM
0.01 0.03 0.02 0.03 0.02 0.03 0.02 0.03 0.18 0.03 0.23 0.03
cluster mass) for the cluster populations presented in Fig. 2. The
additional cooling by dust leads to an additional condensation of gas
relative to the cooling model of about 22, 15 and 4 per cent (median
values over the displayed range of mass) for the models D1, D2 and
D4, respectively. The fraction of baryons that condense out of the
ICM in the cooling simulation is about 35 per cent for clusters with
M >2 10
14
h
1
M

and about 45 per cent for systems with lower

mass. We note however that our present simulations do not include
non-gravitational heating mechanisms to regulate the condensation
of gas and therefore overcooling is in practice limited by resolution.
5.2 Evolution of the scaling relations
We now turn to the discussion of the evolution of the cluster scaling
laws in our simulations. Here, we apply the t to a power-law
procedure described in Section 4 to derive the logarithmic slope, ,
of our tting functions, equations (11)(14). As mentioned earlier,
this quantity measures evolution departures relative to the self-
similar expectations for each scaling.
In Figs 47, we plot the redshift dependence of the power-law
slopes, (top panels), and normalizations, log y
0
(middle panels),
for our T
mw
M, SM, YM and L
X
M scalings, respectively. The
bottom panels show straight line best ts, up to z = 1, to the data
points in the middle panels of each gure. The slopes of these lines
are the parameters in equations (11)(14). We decided not to
include data points above z = 1 in the computation of because
cluster numbers drop rapidly (below 20) which, in some cases,
causes large oscillations in the computed normalizations. Moreover
in the case of the L
X
M relation, the evolution of y
0
(z) with redshift
appears to deviate from a straight line above z . 1. In Table 2,
we provide a complete list of the log A, and tting parameters
and associated statistical errors for all scalings and cooling models
investigated in this paper. The displayed values are valid in the
redshift range 0 <z <1. In the top and middle panels, the coloured
bands correspond to the 1 envelope of the best-tting errors
obtained at each redshift for and log y
0
. The shaded areas in the
bottom panels are rms t dispersions of the normalizations, log y
0
,
computed for the cooling model using equation (17).
C _
2009 The Authors. Journal compilation
C _
2009 RAS, MNRAS 396, 849859
856 A. C. da Silva et al.
Figure 4. Evolution of the slope (top panel), normalization (middle panel)
and normalization best-tting lines (bottom panel) of the T
mw
M cluster
scaling relation for the C (triangles, solid line), D4 (diamonds, triple-dot
dashed line), D3 (squares, short-dashed line), D2 (circles, dashed line) and
D1 (crosses, dotdashed line) simulation models. Colour bands are best-
tting errors to the cluster distributions at each redshift. The shaded area
in the bottom panel is the rms dispersion of the normalization t for the
cooling model (see text for details).
Results from different simulation runs are coded in the following
way: triangles and solid lines stand for the cooling model, dia-
monds and triple-dotdashed lines represent the D4 model, squares
and short-dashed lines are for the D3 model, circles and dashed
lines are for the D2 models and crosses and dotdashed lines are
for the D1 model. Here, we have omitted the D5 model for clarity.
It provides the same tting results as the cooling model. This con-
rms expectations and the comments made in the last paragraph of
Section 2.3.
The top panels of these gures allow us to conclude that the
slopes of our scalings are fairly insensitive to dust cooling. These
also show no evidence of systematic variations with redshift for all
scalings, which is an important requirement when tting the cluster
distributions with power laws of the form equations (11)(14). The
redshift independence of the slopes with the dust model conrms our
ndings at redshift zero. The scatter (non-systematic oscillations)
at high redshift is caused by the decrease of the number of clusters
with M
lim
5 10
13
h
1
M

, the sample selection used for all ts.

The main effect of cooling by dust is reected in the changes it
produces in the normalizations of the cluster scaling laws. Again,
the impact of dust is different depending on the scaling under con-
sideration. For the T
mw
M scaling in Fig. 4, we see a systematic
variation with the dust model (ordered in the following way: C,
D4, D3, D2, D1), but differences between models are within the
Figure 5. Evolution of the slope (top panel), normalization (middle panel)
and normalization best-tting lines (bottompanel) of the SMcluster scaling
relation for the C, D4, D3, D2 and D1 simulation models. Symbols, lines
and colours are the same as in Fig. 4.
errors and t dispersions of each other. For the evolution of the nor-
malizations of the SM, YM and L
X
M scalings (see Figs 57),
we conclude that the inclusion of dust cooling causes signicant
departures from the standard radiative cooling model depending
on the dust model parameters. For example, this is clear from the
non-overlapping errors and t dispersions of the normalizations for
the D2 and D1 models. For all scalings, the relative strength of the
effect of dust follows the relative intensity of the cooling functions
presented in Section 2.3. This orders the models in the following
way: C, D4, D3, D2, D1 with increasing normalizations for the
T
mw
M and SM scalings and decreasing normalizations for the
YM and L
X
M relations.
We end this section by noting that we nd positive evolution
[relative to the expected self-similar evolution, i.e. for a given x in
equation (15) the property yf is higher at higher redshifts] for the
YM and L
X
M (models D1 and D2 only) relations, whereas
the T
mw
M and SM relations show negative evolution relative to
the self-similar model. This is in line with the ndings from sim-
ulations with radiative cooling of similar size and cosmology (see,
e.g., da Silva et al. 2004; Aghanim et al. 2009).
6 DI SCUSSI ON
6.1 Efciency of the dust cooling
In agreement with the cooling functions of Montier &Giard (2004),
the dust cooling is most effective in the cluster temperature regime.
The relative importance of the dust cooling with respect to the gas
C _
2009 The Authors. Journal compilation
C _
2009 RAS, MNRAS 396, 849859
The impact of dust on cluster scaling laws 857
Figure 6. Evolution of the slope (top panel), normalization (middle panel)
and normalization best-tting lines (bottompanel) of the YMcluster scaling
relation for the C, D4, D3, D2 and D1 simulation models. Symbols, lines
and colours are the same as in Fig. 4.
radiative cooling is strongly dependent on the dust abundances and
the intrinsic physical properties of the dust. This is clearly shown
in our scaling relations results.
(i) The T
mw
M relation is almost unchanged when adding dust
cooling to the radiative gas cooling (see Fig. 4). Our results show
that the (mass weighted) temperaturemass relation within R
200
is
essentially driven by the gravitational heating of the gas (due to its
infall on the cluster potential well), and that the physics of baryons
(at least for the physics implemented in the simulations presented
in this paper) plays very little role in the outer parts of haloes which
dominate the estimation of the mass-weighted temperature and in-
tegrated mass. Since gas cooling tends to disturb the dark matter
distribution at the centre of clusters in high resolution simulations
(Gnedin et al. 2004), the cooling by dust may amplify this effect,
and thus modify scaling relations like the T
mw
M. In the case of
observationally derived quantities, scaling laws will be drawn from
overall quantities that will proceed from mixed-projected informa-
tion over a wide range of radii. If a gradient exists in the dust effect
towards the cluster centre, an overall temperature might bear the
signature of the structural effect of dust. Anyway this quantica-
tion is beyond the scope of this paper and will be investigated in a
forthcoming paper. There is also no signicant effect between the
different dust models and the radiative case on the evolution of the
slope and normalization of the T
mw
M relation.
(ii) On the other hand, the other three scaling laws are deeply
related to the baryonic component of the clusters. The clear ef-
fect on the SM, YM and L
X
M relations illustrates this fact (see
Figure 7. Evolution of the slope (top panel), normalization (middle panel)
and normalization best-tting lines (bottom panel) of the L
X
M cluster
scaling relation for the C, D4, D3, D2 and D1 simulation models. Symbols,
lines and colours are the same as in Fig. 4.
Figs 57). We found that the slopes of these scalings remain fairly in-
sensitive to dust, whereas normalizations show signicant changes
depending on the dust parameters. Relative changes in the normal-
izations at redshift zero and M
0
= 10
14
h
1
M

can be as high as
25 per cent for L
X
M and 10 per cent for the SM, YM relations for
the D1 model. Models with lower dust abundances and MRN grain
size distributions present smaller but systematic variations relative
to the C model. As any other cooling process, the cooling due to
dust tends to lower the normalizations of the YM and L
X
M scal-
ings due to the decrease of the hot gas fractions and densities which
dominate the increase of temperature. The increase of normaliza-
tions for the SM and T
mw
M relations with added dust cooling is
also in line with expectations because cooling converts cold, dense
gas into collisionless star-forming material, which raises the mean
temperature and entropy of the remaining gas.
(iii) Our simulations allow us to quantify the relative impact of
the dust parameters on the investigated cluster scalings (see Figs 4
7 and Table 2). From one model to another, one can identify two
clear effects due to dust. (1) It shows the expected effect of the dust
abundance, which frommodels D4 to D2 increases by a factor of 10,
producing a change of normalizations relative to the purely radiative
case (model C), from almost 0 per cent contribution to about 14, 5
and 6 per cent contributions for the L
X
M, YM and SM relations,
respectively. (2) Even more striking is the effect of the intrinsic dust
grain physical properties. The variation of normalizations relative
to the C model changes from a 0 per cent level for model D4 to
about 25 per cent (L
X
M) and 10 per cent (YMand SM) for model
D1 (i.e. the relative change from models D2 to D1 is about 13 and
C _
2009 The Authors. Journal compilation
C _
2009 RAS, MNRAS 396, 849859
858 A. C. da Silva et al.
5 per cent, respectively). All these percentages were calculated using
normalizations at redshift zero and M
0
= 10
14
h
1
M

. Therefore,
the size of the grains becomes an equally important parameter vary-
ing the efciency of the dust cooling. The smaller the grain, the
stronger the cooling.
(iv) From Figs 47 one nds that differences between normal-
izations become progressively important with decreasing redshift.
This conrms expectations because metallicity was modelled in
simulations as a linearly increasing function of time. Although our
implementation of metallicity should only be regarded as a rst
order approximation to the modelling of more complex physical
processes (acting on scales below the resolution scale of the present
set of simulations), it would be interesting to investigate whether
a similar effect remains (i.e. the effects of dust become progres-
sively important at low redshift) when such processes are taken into
account throughout the formation history of galaxy clusters (see
discussion below).
6.2 Limitation of the dust implementation
6.2.1 Injection and destruction of dust in the ICM
In order to implement the presence of dust in the numerical simula-
tions, we chose a zero-order approach: we directly correlated the
presence of dust with the presence of metals under the assumption
that there is no segregation in the nature of the material withdrawn
from galaxies and injected into the IGM/ICM (metals, gas, stars or
dust). However, this assumption suffers from limitations linked to
the dust lifetime. Indeed, dust suffers sputtering whereas metals re-
main (i.e. are not destroyed) in the IGM/ICM. As detailed in Draine
& Salpeter (1979) and Montier & Giard (2004), the dust lifetime
before sputtering for a thermal plasma (with 10
6
< T < 10
9
K) can
be expressed as

sput
= 2 10
7

10
3
cm
3
n
H

a
0.01 m

yr. (18)
Typically, for densities met in the core of dense clusters, i.e. n
H

10
3
cm
3
, and grain sizes ranging from 0.001 to 0.5 m, the dust
grains have lifetimes of 10
6
< t
dust
< 10
9
yr. From equation (18),
we see that big grains have longer lifetimes than small grains. Al-
though small grains are destroyed quickly, their population is con-
tinuously replenished by the sputtering of bigger grains as long as
dust keeps being injected into the ICM. The overall efciency of
the cooling by dust in the ICM/IGM is thus strongly linked to the
physical processes of dust injection (i.e. enrichment) in the ICM,
but also to the medium thermodynamical properties (i.e. temper-
ature and density), the dust grains being destroyed more quickly
in denser regions (i.e. the cluster core) where the cooling by dust
is also the most efcient. Whether the processes of dust injec-
tion replenish the population of the dust grains and balance its
destruction by sputtering over large enough time-scales for the dust
cooling to be efcient is a question out of reach of our current set
of simulations. Indeed, our simulations do not model the physical
processes of dust creation and destruction. This will be a further
level of development in the implementation of the dust physics
in future simulations. Assuming that a full galaxy mass of stars
and gas is dispersed to the ICM within 10
9
yr, and that the dust
lifetime in the ICM is of the order of
cl
sput
= 10
8
yr, from equa-
tion (4) we can derive a crude estimation for the f
d
parameter: f
d
=
Z
cl
d
/Z
gal
d
=(M
gal
tot
/M
cl
gas
) (
gal
/
cl
sput
) .(1/10) (10
8
/10
9
) .0.01.
As for the present observational limits, this crude estimate falls
within the value of f
d
we have tested.
Therefore our whole analysis is to be considered within the frame-
work of our basic implementation of the effects of dust with the
objective of assessing whether dust has a signicant impact on
large-scale structure formation, and consequently to quantify these
effects at rst order. However, making use of the cooling function by
Montier & Giard (2004), we have performed a fully self-consistent
implementation of the effect of dust as a cooling vector of the
ICM/IGM. Indeed, on the computation of the cooling function, the
implementation encapsulates the major physical processes to which
dust is subjected and acts as a non-gravitational process at the scale
of the ICM and the IGM.
6.2.2 Correlation of the dust and metal abundances
As already mentioned, we directly correlated the abundance of dust
with metallicity, thus to the metallicity evolution, the chosen evolu-
tion law of which is quite drastic: Z = 0.3(t/t
0
) Z

. Indeed, if the
metallicity at z = 0 is normalized to the value of 0.3 Z

, it is low-
ered to 0.2 at z =0.5 and 0.1 at z =1. However, other numerical
works based on simulations including physical implementation of
metal enrichment processes but without dust agree well with obser-
vational constraints (mainly provided by X-ray observations of the
Fe K line) which indicate high metallicity values, Z 0.3 Z

, up
to redshifts above 1.0 (Borgani et al. 2008; Cora et al. 2008). This
shows that, as for the stellar component which is already in place in
galaxies when clusters form, the metal enrichment of the ICM/IGM
has occurred through the feedback of galaxy formation and evolu-
tion, and therefore it de facto strongly enriched the IGM/ICMbelow
z = 1. It also might give hints that the high metallicity of clusters
could be correlated to the dust enrichment of the IGM/ICM. Indeed,
the amount of gaseous iron in galaxies such as the Milky Way is
0.01 Z

. An early enrichment of dust in the IGMand/or the ICM,

which once sputtered will provide metals, could explain part of the
iron abundances found in the ICM at low redshifts. This hypothesis
seems to be consolidated by a few works that have investigated dust
as a source for metals in the material stripped from galaxies via
dynamical removal within already formed clusters (Aguirre et al.
2001) or via an early IGM enrichment at high redshift during the
peak of star formation around z = 3 (Bianchi & Ferrara 2005). The
latter work stressed that only big grains (a > 0.1 m) can be trans-
ported on a few 100 kpc physical scale, however leading to a very
inhomogeneous spatial enrichment in metals once the dust grains
are sputtered. For all these reasons, by underestimating the metal-
licity at high redshifts, we might have underestimated the amount
of dust injected into the ICMat high redshift, and thus the efciency
of dust cooling when integrated from early epochs down to redshift
zero.
7 CONCLUSI ON
In this work, we have presented the rst simulations of structure
formation investigating the effect of dust cooling on the properties
of the ICM. We have compared simulations with radiative plus dust
cooling with respect to a purely radiative cooling simulation. We
have shown the following.
(i) The cooling due to dust is effective in the cluster regime and
has a signicant effect on the baryon driven statistical properties
of clusters such as L
X
M, YM, SM scaling relations. As an added
non-gravitational cooling process dust changes the normalization
of these laws by a factor up to 25 per cent for the L
X
M relation,
and up to 10 per cent for the YM and SM relations. In contrast,
C _
2009 The Authors. Journal compilation
C _
2009 RAS, MNRAS 396, 849859
The impact of dust on cluster scaling laws 859
dust has almost no effect on a dark matter driven scaling relation
such as the T
mw
M relation.
(ii) The inclusion of cooling by dust does not change signicantly
the slopes of the cluster scaling laws investigated in this paper. They
compare with the results obtained in the radiative cooling simulation
model.
(iii) Through the implementation of our different dust models,
we have demonstrated that the dust cooling effect at the scale of
clusters depends strongly on the dust abundance in the ICM, but
also to a similar extent on the size distribution of the dust grains.
Therefore, the dust efciency is strongly dependent on the nature of
the stripped and ejected galactic material, as well as the history of
these injection and destruction processes along the cluster history.
Indeed, the early enrichment of dust might provide an already mod-
ied thermodynamical setup for the to-be-accreted gas at lower
redshifts.
The setup of our simulations and the limitation of our dust im-
plementation can be considered a zero-order test with which we
demonstrated the active effect of dust on structure formation and
especially at cluster scales. In order to go one step further, a per-
spective of this work will be needed to couple the radiative cooling
function of dust with a physical and dynamical implementation of
the creation and destruction processes of dust in the IGM/ICM.
ACKNOWLEDGMENTS
The authors thank the referee for his discerning and fair comments.
We are deeply indebted to Peter Thomas, Orrarujee Muanwong and
collaborators for the their part in writing the original Sussex cluster
extraction software used in this work, and to Nabila Aghanim for
discussions and providing us access to the IAS (Orsay) computing
facilities where simulations were run. We thank Mauro Roncarelli,
Alain Blanchard, Peter Thomas and Nabila Aghanim for fruitful
discussions and comments on the manuscript. EP and LM ac-
knowledge the support of grant ANR-06-JCJC-0141. JL beneted
from the support of the French-Portugese Luso actions (PAUILF
PI: Alain Blanchard & Pedro Viana). AdS acknowledges support
from Funda c ao Ci encia e Tecnologia (FCT) under the contracts
SFRH/BPD/20583/2004 and CI

ENCIA 2007.
REFERENCES
Aghanim N., da Silva A. C., Nunes N., 2009, A&A, 496, 637
Aguirre A., Hernquist L., Schaye J., Katz N., Weinberg D., Gardner J., 2001,
ApJ, 561, 521
Arnaud M., 2005, Background Microwave Radiation and Intracluster Cos-
mology. IOS Press, the Netherlands, and Societ` a Italiana di Fisica,
Bologna, Italy, p. 77
Bai L., Rieke G., Rieke M., Hinz J., Kelly D., Blaylock M., 2006, ApJ, 639,
827
Bai L. et al., 2007, ApJ, 664, 181
Bardeen J. M., Bond J. R., Kaiser N., Szalay A. S., 1986, ApJ, 304, 15
Bianchi S., Ferrara A., 2005, MNRAS, 358, 379
Borgani S., Fabjan D., Tornatore L., Schindler S., Dolag K., Diaferio A.,
2008, Space Sci. Rev., 134, 379
Bovy J., Hogg D., Moustakas J., 2008, ApJ, 688, 198
Cattaneo A., Teyssier R., 2007, MNRAS, 376, 1547
Chelouche D., Koester B. P., Bowen D. V., 2007, ApJ, 671, L97
Cora S. A., Tornatore L., Tozzi P., Dolag K., 2008, MNRAS, 386, 96
Conroy C., Ostriker J. P., 2008, ApJ, 681, 151
Conroy C., Wechsler R. H., Kravtsov A. V., 2007, ApJ, 668, 826
Couchman H. M. P., 1991, ApJ, 368, L23
Couchman H. M. P., Thomas P. A., Pearce F. R., 1995, MNRAS, 452, 797
da Silva A. C., Kay S. T., Liddle A. R., Thomas P. A., 2004, MNRAS, 348,
1401
Domainko W. et al., 2006, A&A, 452, 795
Draine B. T., Salpeter E. E., 1979, ApJ, 231, 77
Dwek E., 1981, ApJ, 247, 614
Dwek E., Rephaeli Y., Mather J. C., 1990, ApJ, 350, 104
Eke V. R., Navarro J. F., Frenk C. S., 1998, ApJ, 503, 569
Giard M., Montier L., Pointecouteau E., Simmat E., 2008, A&A, 490, 547
Gnedin O. Y., Kravtsov A. V., Klypin A. A., Nagai D., 2004, ApJ, 616, 16
Kaiser N., 1986, MNRAS, 222, 323
Kapferer W. et al., 2006, A&A, 447, 827
Krick J. E., Bernstein R. A., 2007, AJ, 134, 466
Loewenstein M., 2006, ApJ, 648, 230
Mathis J. S., Rumpl W., Nordsieck K. H., 1977, ApJ, 217, 425
McNamara B. R., Nulsen P. E. J., 2007, ARA&A, 45, 117
Moll R. et al., 2007, A&A, 463, 513
Montier L., Giard M., 2004, A&A, 417, 401
Montier L., Giard M., 2005, A&A, 439, 35
Muanwong O., Thomas P. A.Kay S. T., Pearce F. R., Couchman H. M. P.,
2001, ApJ, 552, L27
Muller S., Wu S. Y., Hsieh B. C., Gonzalez R., Loinard L., Yee H., Gladders
M., 2008, ApJ, 680, 975
Murante G. et al., 2004, ApJ, 607, L83
Murante G., Giovalli M., Gerhard O., Arnaboldi M., Borgani S., Dolag K.,
2007, MNRAS, 377, 2
Pearce F. R., Couchman H. M. P., 1997, New Astron., 2, 411
Pearce F. R., Thomas P. A., Couchman H. M. P., Edge A. C., 2000, MNRAS,
317, 1029
Popescu C., Tuffs R., Fischera J., V olk H., 2000, A&A, 354, 480
Sarazin C. L., 1988, Cambridge Astrophysics Series. Cambridge Univ. Press,
Cambridge
Springel V., Hernquist L., 2003, MNRAS, 339, 312
Stickel M., Lemke D., Mattila K., Haikala L. K., Klaas M., 1998, A&A,
329, 55
Stickel M., Klaas U., Lemke D., Mattila K., 2002, A&A, 383, 367
Sugiyama N., 1995, ApJS, 100, 281
Sutherland R. S., Dopita M. A., 1993, ApJS, 88, 253
Thacker R. J., Couchman H. M. P., 2000, ApJ, 545, 728
Thomas P. A., Couchman H. M. P., 1992, MNRAS, 257, 11
Thomas P. A. et al. (the Virgo Consortium), 1998, MNRAS, 296, 1061
Voit G. M., 2005, Rev. Mod. Phys., 77, 207
Weingartner J., Draine B. T., Barr D., 2006, ApJ, 645, 1188
This paper has been typeset from a T
E
X/L
A
T
E
X le prepared by the author.
C _
2009 The Authors. Journal compilation
C _
2009 RAS, MNRAS 396, 849859